Search skills/

OmO (oh-my-openagent) Claude skills

An agent harness and CLI (omo) with multi-model orchestration; its .agents/skills folder holds mostly maintainer workflows for the project itself.

code-yeongyu/oh-my-openagent 60 skills Planning, TDD & debuggingSkill & agent tooling
Our verdict Recommended with caveats

An ambitious agent runtime rather than a skill pack; its skills are project-specific, the headline installer is a curl-to-bash script, and GitHub detects no standard license.

3/5 Depth expertise shipped ! Security Broad access 0d Maintenance Updated today 70K Adoption stars ? License no clear license

The author changed 3 of 60 skill folders after our review (noticed Oct 29, 2026). The review describes the earlier version; we recheck changed skills weekly, the most significant changes first.

What OmO (oh-my-openagent) does

This repository is the home of OmO, a standalone agent command (omo) that the README describes as turning your model subscriptions into finished work through multi-model orchestration, a memory system and a keyword-triggered mode (typing 'ulw' or 'mass ulw' with a prompt). It was formerly an OpenCode plugin and now also ships a Codex edition. The skill count is high, 60 in the metadata, but the ones we could read in .agents/skills are the maintainers' operations tooling: codex-qa runs the Codex plugin in an isolated CODEX_HOME against a mock model, github-triage produces read-only issue and PR reports with permalinks as evidence, get-unpublished-changes drafts release notes by layer, and publish, pre-publish-review, hyperplan, remove-deadcode and security-research cover release and review chores. Several are long, up to 26 KB, with reference files and scripts. They assume this monorepo's layout and are not general-purpose skills. The practical product is the omo binary, installed with a shell installer or from npm as omo-ai, and the harness behavior around it.

Good for

  • People who want a full alternative harness rather than a skill add-on
  • Maintainers studying how a large agent project automates QA and releases

Not for

  • Dropping a few skills into an existing Claude Code setup
  • Anyone who needs a clearly stated open source license

What's inside depth 3/5

Sixty skill folders with some long, reference-backed ones and a few dozen scripts, but they are repo-maintenance workflows with almost no tests or evals.

Skills
60
Reference docs
227
Scripts
344
Tests / evals
199
Templates
15
Avg SKILL.md
14.8 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Model subscriptions or API access for the providers you use (Claude, ChatGPT, Kimi, GLM)
  • Bun or npm if you install the omo-ai package
  • Telemetry: Anonymous PostHog lifecycle events appear in the Codex edition's hooks; the project's QA docs show OMO_DISABLE_POSTHOG=1 and OMO_CODEX_DISABLE_POSTHOG=1 to turn them off. The /bug command shows the report before sending it.

Security review

Broad access

This skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.

  • Uses your browser cookies or logged-in profileThe browser skill's default 'attached' engine drives the user's own signed-in browser and open tabs through BrowserSkill (extension registered in the browser by browser-install.mjs); it also offers a 'cloak profile' with bot-scoring evasion.packages/shared-skills/skills/browser/SKILL.md:8
  • Pushes or merges without askingwork-with-pr pushes the branch, opens a PR and by default arms auto-merge (gh pr merge --merge --auto --delete-branch) 'unless the user explicitly told you not to merge'; no per-merge confirmation step..agents/skills/work-with-pr/SKILL.md:299

The oh-my-openagent / omo harness and its skills. The install path is a vendor curl | bash, the Codex plugin runs hooks on every session and prompt (including PostHog telemetry and auto-update, with documented opt-out env vars), and the browser skill drives your own signed-in browser through a registered extension. Maintainer skills such as work-with-pr push branches and arm auto-merge by default. We hold under 2% of the repository, so this is a partial review.

Worth knowing

  • Installs with curl | shHeadline install is curl -fsSL https://get.omo.dev/install.sh | bash (vendor host; README says it verifies release checksums, but the script is not in our snapshot).README.md:50
  • Sends usage data by default (can be turned off)The Codex plugin's telemetry component sends a PostHog event at SessionStart by default; OMO_DISABLE_POSTHOG=1 / OMO_CODEX_DISABLE_POSTHOG=1 turn it off (.agents/skills/codex-qa/references/install-verify.md:11)..agents/skills/codex-qa/references/components-hooks.md:24
  • installs_hooksThe omo-codex plugin wires hooks (SessionStart, UserPromptSubmit, PreToolUse, PostToolUse, Stop, PostCompact) for rules, ultrawork, LSP, comment-checker, telemetry, bootstrap and auto-update; they run on every session/prompt/tool call once the plugin is installed. The hooks.json itself is not in our snapshot, only this reference..agents/skills/codex-qa/references/components-hooks.md:3
  • Changes files outside your projectBrowser onboarding installs the BrowserSkill CLI/daemon and registers its extension in the user's browser; the onboarding skill can import API keys/MCP servers/skills from OpenCode and other agents into ~/.omo (user opt-in lane).packages/shared-skills/skills/browser/scripts/browser-install.mjs:12
  • Installs packages or toolsinstall.sh tries brew, npm install -g @ast-grep/cli, cargo install, pip/pipx, nix, mise, then downloads a GitHub release zip (pinned 0.45.0, no checksum check) into ~/.cache/omo-ast-grep/bin; install.ps1 does the same on Windows.packages/shared-skills/skills/ast-grep/install.sh:7
  • Runs remote code at an unpinned versionPackage-manager fallbacks install ast-grep-cli / @ast-grep/cli without a version pin.packages/shared-skills/skills/ast-grep/install.sh:72
  • Relies on a CLI or package we could not reviewSkills rely on the omowright library, BrowserSkill, the omo/opencode/codex CLIs, ast-grep, gh, tmux and Docker, none of which are fully present in our snapshot.packages/shared-skills/skills/browser/scripts/omowright.mjs:1
  • Makes network requestsgithub-triage runs read-only gh CLI queries (the SKILL.md forbids comment/merge/close actions); QA scripts talk to local loopback servers and mock models only..agents/skills/github-triage/scripts/gh_fetch.py:60

What we checked

The repository (12004 files) on Oct 29, 2026. Not a full review: 11839 of 12004 files were not downloaded (large repository; we reviewed the skill folders); it contains compiled binaries or archives we cannot read; we read 5 of 47 code files; it relies on code outside the repository: 11,839 of 12,004 repo files are not in our snapshot, including the omo/opencode plugin source, packages/omo-codex/plugin/hooks/hooks.json, the get.omo.dev installer, and most SKILL.md files; https://get.omo.dev/install.sh and the omo-ai npm package; omowright library and BrowserSkill CLI/extension (staged at runtime), PostHog telemetry endpoint.

NVIDIA SkillSpector static scan: 85 pattern hits (Excessive Agency 19, Tool Misuse 16, Agent Snooping 13, Data Exfiltration 12, Rogue Agent 6, Supply Chain 4). SkillSpector hits are mostly doc/command mentions inside QA and ops skills (rm of temp dirs and the omo binary, DELETE routes in the opencode server docs, ~/.codex/config.toml snapshot checks proving QA does not touch the real config), plus the real curl|bash installer and the PostHog telemetry hooks. The 'concealed executable' hits are duplicate gh_fetch.py copies in .agents and .opencode.

A review, not a certification. Skills change after we check them — read the files before installing.

License

GitHub reports no recognized license (NOASSERTION); check the repository's LICENSE file before commercial use.

Can it run unattended?

Yes. The README and onboarding skill describe an ultrawork mode with continuation and loop hooks for long jobs, though we did not verify their stop conditions.

Alternatives to OmO (oh-my-openagent)

wshobson/agents

wshobson/agents

A multi-harness plugin marketplace with 94 plugins, 202 agents, 185 skills and 105 commands for Claude Code, Codex, Cursor, OpenCode, Copilot, Antigravity and Pi.

With caveats Check first Depth 2/5

Orca

stablyai/orca

An open-source desktop and mobile app for running several coding agents in parallel git worktrees, plus 11 thin skills that hand agents the app's CLI.

With caveats Check first Depth 2/5

herdr

herdrdev/herdr

A Rust terminal multiplexer for coding agents, with a skill that lets one agent split panes, start other agents and wait on them.

With caveats Check first Depth 2/5

Claude Octopus

nyldn/claude-octopus

An escalation layer that runs tasks past up to twelve external AI providers (Codex, Copilot, Ollama, Grok, more) with a consensus gate, plus 63 skills.

With caveats Check first Depth 4/5

FAQ

What is oh-my-openagent?

It is the repo behind OmO, a coding-agent harness started as an OpenCode plugin. It now provides an omo command with multi-model orchestration and memory, and a Codex edition.

How do I install OmO?

The README gives a curl-based installer (curl -fsSL https://get.omo.dev/install.sh | bash) and a PowerShell equivalent, or you can install the omo-ai package with bun or npm. Read the installer first if you prefer.

Does OmO collect telemetry?

The Codex edition includes an anonymous PostHog telemetry hook. The project's own QA documentation sets OMO_DISABLE_POSTHOG=1 to disable it, so use that variable if you want it off.

Can I use the skills in oh-my-openagent without the harness?

Mostly not usefully. The skills under .agents/skills are written for this monorepo's releases, QA and triage and refer to its packages and paths.

Sources: GitHub repository (stars, license, file tree, last push Oct 29, 2026). Review based on the README and SKILL.md files as published.